Matrix Multiplication- Handling Different Dimensions

Matrix Multiplication: The Dimension Problem Nobody Warns You About

You learned matrix multiplication in school. You multiplied rows by columns. Easy. Then you tried to multiply two matrices and got hit with a dimension mismatch error.

That's not a bug. It's the rule.

Matrix multiplication isn't defined for arbitrary dimensions. The columns of the first matrix must match the rows of the second. If they don't, you can't multiply them. Period.

The One Rule That Determines Everything

For matrix multiplication to work, the number of columns in matrix A must equal the number of rows in matrix B.

If A is (m × n) and B is (p × q), you need n = p.

The result will be a matrix of size (m × q).

Visual Breakdown

Think of it this way: the inner dimensions must match. The outer dimensions give you the result size.

(m × n) × (n × q) = (m × q)

See how n appears in both? That's the connection point.

Dimension Examples That Actually Make Sense

Example 1: Standard Square Multiplication

A is (2 × 3). B is (3 × 2).

Columns of A = 3. Rows of B = 3. They match.

Result: (2 × 2)

Example 2: Non-Square Works Fine

A is (3 × 4). B is (4 × 2).

4 = 4. It works.

Result: (3 × 2)

Example 3: This Won't Work

A is (2 × 3). B is (2 × 3).

Columns of A = 3. Rows of B = 2. 3 ≠ 2.

You cannot multiply these. No workaround. Pick different matrices.

Example 4: Vector Multiplication

A is (1 × 3). B is (3 × 1).

3 = 3. It works.

Result: (1 × 1) — a single number, also called a scalar.

Common Mistakes That Will Waste Your Time

• Assuming commutativity doesn't exist. A × B ≠ B × A in matrix multiplication. The dimensions might not even allow B × A.
• Forgetting the order matters. If A × B works, B × A might not even be defined.
• Confusing element-wise with matrix multiplication. Element-wise multiplication requires same dimensions. Matrix multiplication requires the matching rule above.
• Transposing when you should be rethinking your approach. Transposing swaps dimensions, which sometimes fixes problems, but it changes the operation.

The Multiplication Process Step-by-Step

Given A (2 × 3) and B (3 × 2), here's what you actually do:

Element (i,j) in the result = row i of A dotted with column j of B.

For result[1][1]: take row 1 of A, column 1 of B, multiply matching elements, add them up.

Repeat for every position in the result matrix.

Tools for Handling Matrix Dimensions

Getting Started: Code Examples

Python with NumPy

NumPy catches dimension errors and tells you exactly what's wrong.

import numpy as np

A = np.array([[1, 2, 3], [4, 5, 6]]) # (2 × 3)

B = np.array([[1, 2], [3, 4], [5, 6]]) # (3 × 2)

C = np.dot(A, B) # Result: (2 × 2)

If dimensions don't match, NumPy throws a clear error message.

JavaScript with math.js

const math = require('mathjs');

const A = [[1, 2, 3], [4, 5, 6]];

const B = [[1, 2], [3, 4], [5, 6]];

const C = math.multiply(A, B);

When Dimensions Don't Match: Your Options

If you need to multiply matrices with mismatched dimensions:

• Transpose one matrix. This flips dimensions. Aᵀ changes (m × n) to (n × m).
• Add padding rows/columns. Zero-padding is common in some applications.
• Reconsider your approach. Maybe matrix multiplication isn't the right operation for your use case.
• Use element-wise multiplication. Requires same dimensions, but sometimes solves the actual problem.

Real-World Dimension Examples

Neural networks use specific dimension rules. Input (batch × features) × Weight (features × neurons) = Output (batch × neurons). The inner dimension (features) must match.

Image processing: applying filters requires careful dimension management. Convolution operations have their own rules that build on basic matrix multiplication.

Computer graphics: transformation matrices multiply in sequence. Each step must satisfy the dimension requirements or the transformation fails.

The Bottom Line

Matrix multiplication with different dimensions isn't complicated. The rule is simple: inner dimensions must match. The result size comes from the outer dimensions.

If your matrices don't fit the rule, you have three choices: transpose, change your matrices, or find a different operation.

No amount of clever coding bypasses this requirement. It's fundamental to the mathematics.